I think Steve Ballmer's quote was something like "Measure a software project's progress by increase in lines-of-code is like measuring an airplane project's progress by increase in weight."
I have been reading the book “apple in China” after hearing the author on a podcast. It has fundamentally altered my view of apple as a company. From a consumer perspective, I thought it was a an amazing company. But looking behind the scenes, I came to understand how morally compromised it has been for a very long time. In retrospect, I feel complicit in things I didn’t understand I was part of.
Anything looks worse when you see behind the curtain. The question is in comparison - who produces technology you want without that behind the scenes behavior (or being dependent on someone else's behind the scenes behavior!)?
I read your comment before reading the article, and I disagree. Maybe it is because I am less actively involved in AI development, but I thought it was an interesting experiment, and documented with an appropriate level of detail.
The section on the identity crisis was particularly interesting.
Mainly, it left me with more questions. In particular, I would have been really interested to experiment with having a trusted human in the loop to provide feedback and monitor progress. Realistically, it seems like these systems would be grown that way.
I once read an article about a guy who had purchased a subway franchise, and one of the big conclusions was that running a subway franchise was _boring_. So, I could see someone being eager to delegate the boring tasks of daily business management to an AI at a simple business.
The whole geo-IP space started back with my startup, Digital Envoy, in 1999 (https://news.ycombinator.com/item?id=39734355). The way we went about it was by providing an API to clients but we actually hosted our entire database (encrypted and in a proprietary format) with clients. The reason for this was for latency (back in 1999 we could get about 0.03ms per transaction something that you can't get on any edge delivered service) which was necessary for the types of clients we went after.
The business was very valuable across a lot of industries - gambling, encryption, advertising, security, adult entertainment, etc. - so there was a lot of demand that also helped smooth out the demand up/down cycles. If one market was cold, another was hot. But basically it's a lot of work and a lot of hand-to-hand combat. This is the best way to learn and get passionate customers. Show up, sell to them, and convince them they need you. You'll learn so much by doing this. And don't use the excuse that you're an introvert or not good at selling. If you want to be an entrepreneur, you need to learn and improve. No one is the best at anything on day one - you won't be either. But you'll get there if you keep at it.
Being the absolute best in the market meant that even having much more better funded competitors ($50m+ for competitors against our $12m in funding) meant we tended to win all the time. And before you ask, if I had this to do over again now, I could do this company for a LOT less money given how commoditized things are. I can tell you the time I almost spent $1m on a storage array until I found a cheaper vendor for $250k. Oh, that storage array was for 1TB of storage. So yeah.
Feel free to ask me anything. If there are enough people who have questions and want to do a chat I'd be happy to host a video call and get peppered with whatever questions you might have.
I’m just now really trying to put myself out there and get into sales to further develop our business. We’ve been lucky with some early network sales, but for the next chapter of our business (https://resonancy.io) I need to build the sales engine and I feel very much out of my depth.
Just telling myself its a practice and trying to chip away at it one week at a time.. At this point identifying potential customers and getting meetings is a real challenge. Been trying Apollo and doing cold outreach but is a slow grind. Also trying my best in linkedin..
IMO - and again, remember this is experience from 20+ years ago although I still do a lot of this now - using these kinds of automated shotgun approaches is a mistake, especially early on and especially if you are selling something. We use an automated engine for our podcast agency (https://edgewise.media) but that is a different pitch when it's "come on my podcast and talk about yourself" versus "I want you to spend money for my widget".
What we did was more labor intensive and a function of these kinds of tools not existing (initial dot-com days). We would read the tech press, etc. and identify companies that might have a use for our tech. Then we would email the top person (guessing their email address usually - and most companies use a small number of variations for addresses) with a custom written, but short, tailored pitch to why they could use us and see if they wanted to talk. No pitch deck or attachments. Literally 2-3 sentences. Something that if even someone is going to quickly delete, something in those 2-3 sentences might catch their eye and make them reply instead of delete. We landed customers like Google (who we ended up later suing - that is a different tale), Doubleclick (who ended up being bought be Google), PayPal, Xbox, and a bunch of others through this method.
Happy to chat more. Startups are hard and all of us should help one another out to make the road a little less hard if possible. Keep at it - I know you'll get there. Hard work will get you there, shortcuts (usually) won't.
Did running the geolocation service allow you to see otherwise hidden abuses by ISPs and other service providers ?
What I mean is: did anything ever break or behave badly ... and then when you investigated you discovered "Oh, look at what these people are doing ..." ?
Interesting question. We never saw anything to this extent but did see when mergers happened how dial-in POPs would be merged and reallocated. Also interesting was how many IP addresses existed for North Korea (less than 32 IIRC) versus Antarctica (more than 2k IIRC). Reminder: I've been out of the business since 2005 (stayed on the board until we were acquired in 2007). So my data is quite dated at this point.
Just FYI, it’s definitely not fraud, it is just that clinicians don’t know when they need to specifically document things that they are actually doing. They do things, don’t specifically mention them in the note because they don’t know it matter, then the billing is lower than it should be. But it’s definitely not fraud.
My wife is a doctor at a major university. They are under pressure right now and are looking to increase revenue. Changing the way they document cases can substantially alter the billing outcome. Note that these are not errors, they are omissions of work done in the note that prevents the downstream billing experts from using higher paying codes.
They have been aware for a few years that many clinicians aren’t documenting their work in the best way for billing. The current solution is to have an annual talk given by the one billing expert in their department pointing out where people often lose revenue due to poor documentation.
Not all the doctors attend this talk. There is no internal process for measuring subsequent improvements quantitatively. There are 85 doctors in her group.
Anyway, this is just to say that something automated to help doctors document their work in a billing friendly way seems powerful. But for my wife’s group, the issue doesn’t seem to be denied claims or “errors” per se. More omissions/sub optimal documentation due to lack of knowledge. Or lack of follow through on knowledge which is only occasionally communicated.
Thank you for the insights, I believe the challenge they have is related to CPT coding - not mistakes or errors, but the completeness' of clinical picture fro m the billing/insurance standpoint. A lot of this coding knowledge is tribal and resides in the head of clinicians. We can help, would greatly appreciate an introduction to your wife's colleagues at dmitry.k@wrkdn.com (Dmitry Karpov), thank you!
Thanks, I will ask my wife to ask the guy who gives the talks (their internal expert in billing matters) if he is interested and send him your email. Personally, I really think your solution has potential to help them, in the sense that they don’t have a sustainable/automated process for training clinicians to write notes in the correct way, or to detect issues when the clinicians don’t know or remember to do so. I know that meeting wasn’t perfectly attended, but even if it had been, they add like 8 new people at the beginning of July, and those people won’t attend the billing meeting until the next one happens in a year. The push for more revenue has been within the last month or so (I think maybe due to Trump related cuts to grant funding and concerns about Medicaid), so it’s an issue that is currently on the mind of leadership.
I'm a software engineer who's spent the last two years in the emergency department scribing before med school. I've only worked at a few hospital systems, but honestly baffles me how piecemeal EHR is, with everyone rolling their own customized implementation of Epic. Monitoring and error checking could have been solved a decade ago. And I agree that most doctors do an inadequate job of charting the work they do - I am sure I routinely miss things like critical care time, documenting heart scores, etc. There is sooo much to do and not enough time to not make billing mistakes on a daily basis. Additionally, many providers have limited insight into reimbursement rates (as they vary by insurance).
Our current approach seems to be quarterly newsletters that nudge people into adding something to their charts.
I am not trying to be antagonistic, just earnestly trying to clarify.
The goal here of your wife's hospital is to try to increase revenue and the outcome, either AI assisted or not, is more accurate visit notes which leads to more accurate billing, which would lead to higher costs to the patient for the same medical care.
If that's right, I suppose a truer reflection of the medical care provided is a good unto itself, but I have to say I don't love the outcome as someone who's a patient and not a shareholder (401k notwithstanding).
Thank you for these thoughts. I think billing inefficiencies caused by documentation make both hospitals and insures lose - one will get paid later, one will pay anyway, but the process is costly for patients who finance both, in the end of the day :(
Thanks, this is helpful. I tried Claude Code, and thought it had a lot of potential, but I was on track to spend at least $20/day.
For a tool that radically increases productivity (say 2x), I think it could still make sense for a VC funded startup or an established company (even $100/day or $36k/year is still a lot less than hiring another developer). But for a side project or bootstrap effort, $36k/year obviously significantly increases cash expenses. $100/month does not, however.
So, I'm going to go back and upgrade to Max and try it again. If that keeps my costs to $100/month, thats a really different value proposition.
I hear you. I’m almost 50, and see it coming as well. I highly recommend the book Outlive by Peter Attia. Just found it recently, wish I had read it at 37. It has really good ideas on extending your “health span “.
The host is clearly trying to push the CEO of Cohere on how all this is going to make money (or just be economic). The CEO is confident, but not in a very specific way. There is a great moment where he is like "we did some proof of concepts with 5 users, and they were pretty good, but when you tell CFOs about the running costs for a full user base, its not viable."
What fascinates me about AI right now is that it seems to have very different economics from traditional software/internet/SaaS businesses. Those business scale super-efficiently. They have some initial startup costs (but still relatively low, especially with cloud providers) and low running costs.
With AI, the initial capital costs to build the model are quite high. And, the running costs to handle queries are also quite high. These companies need to find use cases that generate value significantly in excess of those costs. If those use cases are out there, they must either involve really significant productivity improvements, or the costs have to come down a lot, or both.
All that said, I remember going to a talk by Adobe's founders, in which they pointed out that when they introduced Postscript, the first Apple printer that ran it was only viable because of a last minute drop in memory prices, and when they started building Photoshop, you could fit six (6!) digital images on a powerful computer.
So, I see why the investment is happening, but its a high risk investment right now hoping to identify both high-value use cases and significant cost savings simultaneously.
reply